⬆️ Updated __init_subclass__.

This is a great way to do something without overriding metaclass.

Author: rentruewang

README.md

@@ -1,6 +1,6 @@
 # Snippets
 
-Snippets is a repository that stores snippets from languages I want to work on.
+_Snippets is a repository that stores snippets from languages I want to work on, as well as some theories / thoughts that I have on the same topic._
 
 [![Code Formatted](https://github.com/rentruewang/snippets/actions/workflows/format.yaml/badge.svg)](https://github.com/rentruewang/snippets/actions/workflows/format.yaml)
 

python/src/sub.py

@@ -0,0 +1,40 @@
+# Copyright (c) 2024 RenChu Wang - All Rights Reserved
+
+from typing import Any
+
+
+class Super1:
+    @classmethod
+    def __init_subclass__(cls, **kwargs: Any) -> None:
+        print(cls, kwargs)
+
+
+class Sub1(Super1):
+    pass
+
+
+class Sub2(Super1, param="hello", argument=123):
+    pass
+
+
+class Super2:
+    @classmethod
+    def __init_subclass__(cls, required: str) -> None:
+        print(cls, required)
+
+
+# Would fail:
+# TypeError: Super2.__init_subclass__() missing 1 required positional argument: 'required'
+# class Sub3(Super2):
+#     pass
+
+
+class Sub4(Super2, required="yes"):
+    pass
+
+
+if __name__ == "__main__":
+    Sub1()
+    Sub2()
+    # Sub3()
+    Sub4()

thoughts/fundamental.md

@@ -0,0 +1,18 @@
+# Computation fundamentals
+
+There are 2 main components in computing, data (measured by space complexity) and compute (measured by time complexity).
+
+I theorize that
+Data = entities, storage, intrinsically exists
+Compute = mutation on data
+
+Data is more fundamental than computation, because computation are performed on data (data = symbol, compute = instruction in Turning machine).
+
+Think of the types of data as Markov, then a library / app is just moving on a path on this Markov state / graph. This state is exactly a Turing machine’s state as I can ask TM to simulate a function for me.
+
+Even transmitting a packet counts as computation, because it's changing state (location) of a piece of data.
+
+How do abstractions come into play?
+
+- I think it's just types of data and the type of compute (abstract, reasoned with invariance) that can be performed on them. Considering that every function is data -> compute -> data, abstraction in terms of breaking down functions are just composing data into intermediate representation, and abstract interfaces are shared traits of different data + specialized compute for each kind of data.
+- Abstractions are just using computation to transform underlying data into the same mathematic model. Compiler optimizing this away = 0 cost. Some data structures might happen to have a better complexity but those are implementation details.